Hello tsialex , looks like the final release of Mojave have same behaviour with the 7870 card , do you know nay fix or update about it? ..thx
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
MP 1,1-5,1 MacPro5,1: BootROM thread | 144.0.0.0.0
- Thread starter tsialex
- WikiPost WikiPost
- Start date
- Sort by reaction score
- Status
Usually Apple correct this personality problem with GM or .1 release.
The only way to get this working 100% of the time is changing Hamachi with Futomaki (works only with Pitcairn version of HD 7870), some people did this successfully years ago, around 2014/15 if I remember correctly. Maybe you can track them (don't know who anymore) and ask for help?
Last edited:
pitcarn is the code name for the GPU used in the 7850/7870/R9 270/X/D300 and a couple other GPUs (one of the more modern Retina 5K iMacs even use it which should be good for long time driver support for the 7950 which uses the same driver) Apple milked the living daylights out of the GCN1.0 chips LOL
Hamachi is the name of frame-buffer personality for the reference 7950
Futomaki is the name of the frame-buffer personality for the reference 7870.
(and for reference the 7950/7970/R9 280/X and D700 use a GPU codenamed Tahiti)
7870 has two versions, only the Pitcairn works with Futomaki. Tahiti LE don't work. I used the incorrect name to the personality to differentiate.
[doublepost=1537860677][/doublepost]
BTW, only the Pitcairn version with Futomaki personality has HDMI correctly working.
I would just state that distinction otherwise someones gonna go looking for a frame buffer personality called Pitcarn and get quite lost.
but that does remind me the D500 is Tahiti LE based, and just like how the 7970 used to report D700 OpenGL/OpenCL engine the 7870 Tahiti LE cards would report D500 as the rendering engine
I wonder if thats still the case or if Apple patched that the same time they patched the 7970.
I was thinking about this earlier, maybe we can get DP1.2 support out that…
AFAIK once your in the OS DP1.2 works fine with the Hamachi frame-buffer personality
the DP1.2 bug is looks to be an EFI ROM Bug especially as it exists on other cards too like the GeForce GTX 680.
pretty lazy on the EFI writers part tho as DP1.2 was part of the DP spec back then and even the AMD Radeon 6000 cards support it IIRC.
BTW side note
I was wondering when you rebuild someones BootROM from a .fd file
what do you do for the Firewire GUID/UUID? I have not seen you mention that one before and I noticed in my own efforts that OS X will complain about it if its invalid (found it out when i was working on this https://forums.macrumors.com/threads/going-off-the-firmware-deep-end.2121762/ )
I did some tests trying to find where Apple hid the FW GUID or how it's calculated, talked with @h9826790 about this 2 or 3 weeks ago. Every test I did indicated that is the same case as Ethernet MAC Address. The same BootROM installed into two diff MP5,1 have two diff FW GUIDs.
I know where Apple stored the GUID with 2011/2012 Macs that have ME.
Last edited:
Successfully upgraded to firmware 139 with NVME Support injected (but not upgraded to mojave because of matching graphiccard). I plan to order soon a RX Vega 56 Red Dragon.
Mojave actual installed on a SATA SSD for testing.
Mojave actual installed on a SATA SSD for testing.
Attachments
Last edited:
Now the NDA has lifted, is there anything interesting to share?
I have to check if Apple changed that with the 18A391, maybe it's still over NDA.Now the NDA has lifted, is there anything interesting to share?
Last edited:
I upgraded to 138, then patched NVME using the google doc instructions. After the initial reset to flash the boot rom, the system booted into my HS disk (SATA) that I used for building the rom image.
After the next boot, I reset my PRAM and now the system is stuck in a (boot -> chime -> reset) loop.
Any ideas?
After the next boot, I reset my PRAM and now the system is stuck in a (boot -> chime -> reset) loop.
Any ideas?
Last edited:
Thanks for the info.
So I take it that means I will have to use a Metal capable GPU that has ”Mac EFI” (i.e. boot screen) to install Mojave and get the latest firmware update for the MacPro5,1?
Since I currently only have a Mac version of the Radeon HD 5870 and a ”non-Mac” GTX 1070 I'm out of luck?
This is not entirely true. Someone got it working and even open-sourced the files though till now only the Mac Pro 3,1 works. While the 5,1 loads the GOP EFI already it does not get the frame buffer information required to drive any monitor.
No, these are two separate things. Getting to MP51.0089.B00 is first requirement for Mojave. You will need an EFI GPU for that. Getting to 10.13.6 is second requirement, which I assume should be no issue.
After that, Metal GPU is required for Mojave and this CAN be a separate (compatible) GPU. This will get you onto firmware 138.0.0.0.0 and also allow Mojave install.
NVIDIA Web Drivers are NOT yet available for Mojave. I'd proceed extremely cautiously with non-recommended GPUs until those NVIDIA Web Drivers are available or until more is known.
Read this: https://support.apple.com/en-us/HT208898
Thanks for the info and link!
I'm already at MP51.0089.B00 and 10.13.6. So maybe I can just do the Mojave update from within 10.13.6 using my ”non-Mac” Nvidia GPU and then – once the Nvidia Web Driver that supports Mojave is released – use another Mac to Screen Share into my Mac Pro and install the Web Drivers from that (this is what I've been doing for every point update for MacOS High Sierra).
My concerns are what will happen if something that previously required me to use the Radeon card with ”Mac EFI” happens after installing Mojave. For example that with ”Mac EFI” was required to get the firmware updates (that came during the High Sierra updates) installed in the MacPro5,1.
How will that work down the line with Mojave updates if I don't have a ”Mac EFI" GPU that is Metal compatible?
Edit:
Interesting that some of the cards listed there on the Apple support page shows cards that (supposedly) don't have ”Mac EFI” (or whatever the right definition of it is?). And no mention that they won't give boot screen?
@star-affinity the big thing that got all of this going is that once you're on 138, it doesn't seem to need an EFI card for firmware updates. Alas, that seems to be where Apple stopped with the improvements.
This is why everyone (with concerns) should have access to one of Apple's officially recommended GPUs for Mojave.
This is why I purchased a Sapphire Pulse RX 580 8GB GPU on sale a few weeks back.
This was before Apple officially acknowledged the EVGA/NVIDIA GTX 680 Mac Edition existed.
Always have an EFI GPU available and do not get rid of it.
Currently AMD cards are the only non-EFI officially supported. The two NVIDIA cards Apple officially supports are EFI. There are no NVIDIA Web Drivers for Mojave available at the moment.
- MSI Gaming Radeon RX 560 128-bit 4GB GDRR5
- SAPPHIRE Radeon PULSE RX 580 8GB GDDR5
- SAPPHIRE Radeon HD 7950 Mac Edition
- NVIDIA Quadro K5000 for Mac
- NVIDIA GeForce GTX 680 Mac Edition
That's what I've had (Radeon HD 5870), but since it doesn't support Metal I'm starting to think if I have to do something about the situation…
Time to start searching for a cheap Kepler card.That's what I've had (Radeon HD 5870), but since it doesn't support Metal I'm starting to think if I have to do something about the situation…
I think maybe his point was that the 3,1 is even older than the 4,1/5,1, obviously uses UGA, and a third party bootloader was able to show GOP bootscreens. So there is probably not an insurmountable reason the 4,1/5,1 could not do so also. I realize that there hasn't been success on that front yet, but there have also been very few knowledgable people trying. Just because it hasn't been achieved yet does not mean it's not achievable.
You do NOT need to install 10.14 to get the boot ROM update. Just need to run the installer to trigger it.
Apart from microcode update. The main benefit is now all PCIe 3.0 device can correctly negotiate at PCIe 2.0 link speed automatically in slot 1 and 2.
Which means all AMD GPU now can run at PCIe 2.0 x16, but not PCIe 1.0 x16 (without resistor mod)